Statistical and Texture Features based Breast Cancer Classification from Multi image Modalities using Support Vector Machines with Gaussian Radial Basis Function
Recent developments made in the Image analysis techniques have revolutionized the healthcare systems. A reliable and robust decision making process has taken the healthcare systems to the next level. Cancer disease is one of the most important diseases which have to be focussed at an earlier stage, so as to decrease the mortality rate. Breast cancer is the most dangerous disease which affects the health of women. Different screening modalities are available, yet the rapid decision making is always in under-development stage. Due to the poor quality of the image modalities, the classification accuracy is not high. This paper is a study of Statistical and texture features based breast cancer classification from Multi-Image Modalities using Support Vector Machines with Gaussian radial basis function. The proposed algorithm comprises three phases, namely, preprocessing, feature extraction and the classification process. At first, the visual appearance of the image is improved by preprocessing techniques like median filtering, histogram equalization and the principal component analysis. Statistical and textural features are extracted by applying Local Binary Pattern (LBP) which helped for classifying the instances from multi-classes. Finally, a nonlinear SVM with Gaussian RBF is used for classifying the instances into its relevant classes. With the help of Gaussian RBF kernel operator, the classes of multiple instances were easily improved. The experimental analysis has stated that the better classification accuracy is achieved for multi-image modalities. The results proved that the MRI images have given a better performance index than the other Mammograms and the Ultrasound images.